Contributed Paper
Human Impacts on Regional Avian Diversity
and Abundance
CHRISTOPHER A. LEPCZYK,∗ †∗∗ CURTIS H. FLATHER,‡ VOLKER C. RADELOFF,∗
ANNA M. PIDGEON,∗ ROGER B. HAMMER,§†† AND JIANGUO LIU†
∗
Department of Forest Ecology and Management, University of Wisconsin–Madison, Madison, WI 53706, U.S.A.
†Center for Systems Integration and Sustainability, Department of Fisheries and Wildlife, Michigan State University, East Lansing, MI
48824, U.S.A.
‡USDA Forest Service, Rocky Mountain Research Station, Fort Collins, CO 80526, U.S.A.
§Department of Rural Sociology, University of Wisconsin-Madison, Madison, WI 53706, U.S.A.
Abstract: Patterns of association between humans and biodiversity typically show positive, negative, or
negative quadratic relationships and can be described by 3 hypotheses: biologically rich areas that support
high human population densities co-occur with areas of high biodiversity (productivity); biodiversity decreases
monotonically with increasing human activities (ecosystem stress); and biodiversity peaks at intermediate
levels of human influence (intermediate disturbance). To test these hypotheses, we compared anthropogenic
land cover and housing units, as indices of human influence, with bird species richness and abundance
across the Midwestern United States. We modeled richness of native birds with 12 candidate models of land
cover and housing to evaluate the empirical evidence. To assess which species were responsible for observed
variation in richness, we repeated our model-selection analysis with relative abundance of each native species
as the response and then asked whether natural-history traits were associated with positive, negative, or
mixed responses. Native avian richness was highest where anthropogenic land cover was lowest and housing
units were intermediate based on model-averaged predictions among a confidence set of candidate models.
Eighty-three of 132 species showed some pattern of association with our measures of human influence. Of
these species approximately 40% were negatively associated, approximately 6% were positively associated, and
approximately 7% showed evidence of an intermediate relationship with human influence measures. Naturalhistory traits were not closely related to the direction of the relationship between abundance and human
influence. Nevertheless, pooling species that exhibited any relationship with human influence and comparing
them with unrelated species indicated they were significantly smaller, nested closer to the ground, had shorter
incubation and fledging times, and tended to be altricial. Our results support the ecosystem-stress hypothesis
for the majority of individual species and for overall species diversity when focusing on anthropogenic
land cover. Nevertheless, the great variability in housing units across the land-cover gradient indicates that
an intermediate-disturbance relationship is also supported. Our findings suggest preemptive conservation
action should be taken, whereby areas with little anthropogenic land cover are given conservation priority.
Nevertheless, conservation action should not be limited to pristine landscapes because our results showed that
native avian richness and the relative abundance of many species peaked at intermediate housing densities
and levels of anthropogenic land cover.
Keywords: biodiversity, birds, ecosystem stress, habitat loss, human-dominated landscapes, species–energy
relationship
∗∗ Current address: Department of Natural Resources and Environmental Management, University of Hawai’i at Mānoa, Honolulu, HI 96822,
U.S.A., email lepczyk@hawaii.edu
††Current address: Department of Sociology, Oregon State University, Corvallis, OR 97331, U.S.A.
Paper submitted January 31, 2006; revised manuscript accepted September 3, 2007.
405
Conservation Biology, Volume 22, No. 2, 405–416
C 2008 Society for Conservation Biology
DOI: 10.1111/j.1523-1739.2008.00881.x
Human Impacts on Diversity
406
Impactos Humanos sobre la Diversidad y Abundancia Regional de Aves
Resumen: Los patrones de asociación entre humanos y biodiversidad tı́picamente muestran relaciones
negativas o cuadráticas negativas y pueden ser descritas por 3 hipótesis: áreas biológicamente ricas con densidades humanas altas co-ocurren con áreas de biodiversidad (productividad) alta; la biodiversidad decrece
monotónicamente con el incremento de las actividades humanas (estrés del ecosistema); y la biodiversidad
alcanza picos en niveles intermedios de influencia humana (perturbación intermedia). Para probar estas
hipótesis, comparamos la cobertura de suelo antropogénica y las unidades de vivienda, como ı́ndices de la
influencia humana, con la riqueza y abundancia de especies de aves en el medio oeste de Estados Unidos.
Modelamos la riqueza de aves nativas con 12 modelos de cobertura de suelo y viviendas para evaluar la
evidencia empı́rica. Para evaluar cuales especies eran responsables de la variación observada en la riqueza,
repetimos nuestro análisis de selección de modelos con la abundancia relativa de cada especie nativa como la
respuesta y luego preguntamos si los atributos de la historia natural estaban asociados con las respuestas positivas, negativas o mixtas. La riqueza de aves nativas fue mayor donde la cobertura de suelo antropogénica fue
menor y las unidades de vivienda fueron intermedias con base en predicciones de modelos entre un conjunto
de confianza de modelos posibles. Ochenta y tres de 132 especies mostraron algún patrón de asociación con
nuestras medidas de influencia humana. De estas especies, approximately 40% se asociaron negativamente,
approximately 6% se asociaron positivamente y approximately 7% mostraron evidencias de una relación
intermedia con las medidas de influencia humana. Los atributos de historia natural no se relacionaron
estrechamente con la dirección de la relación entre la abundancia y la influencia humana. Sin embargo, al
combinar especies que mostraron alguna relación con la influencia humana y compararlas con especies no
relacionadas encontramos que eran significativamente menores en tamaño, anidaban más cerca del suelo,
tenı́an tiempos de incubación y salida del nido más cortos y tendı́an a ser altriciales. Nuestros resultados
soportan la hipótesis del estrés del ecosistema para la mayorı́a de las especies individuales y para la diversidad
de especies total al considerar la cobertura de suelo antropogénico. Sin embargo, la gran variabilidad en las
unidades de vivienda en el gradiente de cobertura de suelo indica que también se soporta una relación de
perturbación intermedia. Nuestros resultados sugieren que se deben tomar medidas preventivas de conservación, con lo cual se darı́a prioridad a áreas con baja cobertura de suelo antropogénica. Sin embargo, las
acciones de conservación no se deben limitar a paisajes prı́stinos porque nuestros resultados mostraron que
la riqueza de aves nativas y la abundancia relativa de muchas especies alcanzaron su máximo en densidades
intermedias de densidades de viviendas y niveles de cobertura de suelo antropogénica.
Palabras Clave: biodiversidad, estrés de ecosistema, paisajes dominados por humanos, pérdida de hábitat,
relación especies-energı́a
Introduction
Increased human domination of the Earth’s ecosystems
(Vitousek et al. 1997) and intensifying human land use
(Foley et al. 2005) raise the question of how biodiversity will be affected by these factors. Currently, the observed patterns of association between humans and biodiversity are described by 3 hypotheses (Fig. 1). First,
the productivity hypothesis states that more productive
systems support more species and more people—a covariation manifesting as a positive correlation between
species richness and human population. The positive correlation is thought to be due to productivity gradients
caused by varying energy availability (Gaston 2005), with
more productive landscapes attracting both humans and
other species. Empirical evidence that species richness
increases with human population density in Africa, Europe, and North America (Balmford et al. 2001; Hawkins
et al. 2003; Gaston & Evans 2004; Luck et al. 2004) appears to support this productivity hypothesis. It remains
unclear, however, whether this correlation reflects spatial congruence stemming from selection of high productivity sites by people and wildlife or a causal link, whereby
Conservation Biology
Volume 22, No. 2, 2008
human activities (supplemental feeding, irrigation, plantings) elevate resources and support a more diverse biota
(Rapport et al. 1985).
Second, the ecosystem-stress hypothesis states that humans are detrimental to species diversity because they
remove habitat and resources of most species and thus
predicts a negative relationship between species richness and human influence (Rapport et al. 1985). Evidence that bird and anuran diversity decreases with increases in urbanization (Clergeau et al. 1998; Cam et al.
2000; Genet et al., in Press) supports the ecosystem-stress
hypothesis.
Third, the intermediate-disturbance hypothesis (Connell 1978) states that landscapes under moderate levels of
human impact have higher habitat and resource diversity
compared with pristine or human-dominated landscapes.
This higher resource diversity thus leads to higher species
diversity and forms a negative quadratic relationship (i.e.,
a hump-shaped relationship) between species richness
and human influence (McDonnell & Pickett 1990). Such
patterns have been observed in a number of plant and animal taxa (reviewed in McKinney 2002), typically along
rural-to-urban gradients (Blair 1996).
Lepczyk et al.
407
Species diversity
(Beissinger & Osborne 1982; Sinclair et al. 2002; Pautasso
2007). Considering the importance of the relationship
between human influence and biodiversity for conservation, it is critical to address which of the 3 hypotheses has
the greatest support with methods that control for confounding factors, which was the objective of this study.
Methods
Human influence
Figure 1. Competing hypotheses that explain the
relationships between species diversity and human
influence (i.e., population, houses): productivity (solid
line), intermediate disturbance (dashed line), and
ecosystem stress (dotted line).
Understanding human impacts on species biodiversity
is of critical importance for developing appropriate conservation plans and management guidelines (Ricketts &
Imhoff 2003). Specifically, the type of relationship that
exists between human impacts and species inhabiting
the landscape will govern the conservation and management approaches that can be used. For instance, if species
diversity and human influence exhibit a positive relationship, conservation conflicts are likely to increase because the increased human demand for resources places
species and their habitats at greater risk (Chown et al.
2003). Thus, conservation efforts should focus on areas
where human activities are already high so as to offset impending conflicts between biodiversity protection and
development (Balmford et al. 2001; Ricketts & Imhoff
2003; Carroll et al. 2004). Conversely, a negative relationship between biodiversity and human activities suggests
the focus should be on areas with low anthropogenic disturbance because they harbor greater diversity and may
be more cost-effective to protect (Luck et al. 2004).
Sorting out the alternative explanations and support
for the 3 hypotheses based on the existing literature is
difficult. Previous researchers used disparate methodologies, scales of analysis, and sources of data (Blair 1996;
McKinney 2002; Hope et al. 2003; Luck et al. 2004) to
examine the relationship. Support for the productivity
hypothesis has generally been at regional and continental
scales (Balmford et al. 2001; Chown et al. 2003; Ricketts
& Imhoff 2003), compared with more local-scale support for ecosystem stress and intermediate disturbance
To test the 3 hypotheses, we used 2 complementary measures of human influence (anthropogenic land cover and
housing units) and 2 measures of biodiversity response
(richness and abundance) across a gradient of 408 landscapes (approximately 1200 km2 each) of the midwestern United States (an area [1.2 × 106 km2 ] twice the
size of the Iberian Peninsula; Fig. 2). These landscapes
spanned the entire range of human influence, from pristine to human-dominated, and were spatiotemporally
congruent with bird data derived from long-term abundance records. They also spanned a gradient that varied
from the less-productive grassland and savanna systems
in the western portion of the study area to the moreproductive temperate forests along the Great Lakes and
eastern portion of the study region (Hurlbert & Haskell
2003).
Biodiversity Estimates of Bird Species
Neither richness nor abundance alone can tell the complete story of an area’s biodiversity. Richness provides
only an estimate of the number of unique species present,
whereas abundance provides information on only a single species. Moreover, if biodiversity measures do not
distinguish between native and exotic species, relationships may be quite different. To avoid the possible confounding effects of exotic and range-expanding species
on the pattern observed, we considered only species native to the study region (see Supplementary Material).
We eliminated Ring-necked Pheasants (Phasianus colchicus), Rock Pigeons (Columba livia), European Starlings
(Sturnus vulgaris), House Finches (Carpodacus mexicanus), and House Sparrows (Passer domesticus). Moreover, avian detections that were not identified to species
(e.g., unidentified Empidonax) were also dropped from
the analysis.
Annual raw counts of breeding birds from 1987 to 1997
were obtained from the North American Breeding Bird
Survey (BBS; Sauer et al. 2003). We selected all Midwestern BBS routes that had ≥3 years of acceptable surveys
during the 11-year period centered on 1992 (Fig. 2). A survey was deemed acceptable if it was completed by a competent observer during the peak breeding-season window specified for a particular location and was conducted
within start time, finish time, and weather standards specified by the survey design (Bystrak 1981; Robbins et al.
Conservation Biology
Volume 22, No. 2, 2008
Human Impacts on Diversity
4-Color Artwork
408
Figure 2. The (a) proportion of
anthropogenic land cover and
(b) total number of housing units
present in each survey landscape
in the midwestern United States.
1986). For all analyses, we selected data from the 3 closest
years to 1992. A 3-year mean richness and abundance was
estimated to control for annual variation in bird counts
that was unrelated to our measures of human influence.
Native bird richness was estimated for each year on
each route with COMDYN (Hines et al. 1999) software,
which probabilistically estimates species richness when
not all species are detected. Specifically, COMDYN uses
the jackknife estimator of Burnham and Overton (1979)
to estimate species richness and accounts for heterogeneity in detectability among species (Boulinier et al. 1998).
For bird abundance calculations we selected all bird
species that occurred on at least 30 BBS routes, for a total
of 132 species (Supplementary Material). The number of
individuals detected, like species, can vary as a function
of observer, species, and landscape context (Sauer et al.
1994; Nichols et al. 2000). Nevertheless, the BBS survey
methodology does not currently allow these biases to
be accounted for analytically at the species level. Consequently, we assumed these biases would not appreciably
affect the pattern of spatial covariation between abundance and our measures of human influence.
Measures of Human Influence
We considered anthropogenic land cover and housing units as complementary measures of human influence—
anthropogenic land cover provides a measure of land use,
whereas housing units provide a measure of human activity. We based anthropogenic land cover on a reclassification of the National Land Cover Data (NLCD; Vogelmann
et al. 2001), a 21-class land-cover scheme derived from
Landsat Thematic Mapper satellite data (30-m resolution)
Conservation Biology
Volume 22, No. 2, 2008
from 1992 and 1993. We created a binary map of anthropogenic and nonanthropogenic land cover. Anthropogenic land cover was composed of the following NLCD
land classes: low-intensity residential; high-intensity
residential; commercial/industrial/transportation; quarries/strip mines/gravel pits; orchards/vineyards/other;
pasture/hay; row crops, small grains, fallow; and urban/
recreational grasses. Similarly, nonanthropogenic (i.e.,
natural) land cover was composed of the following NLCD
land classes: transitional, forest (deciduous, evergreen,
mixed), shrubland, grasslands/herbaceous, and wetland
(woody, emergent herbaceous). We determined the number of housing units from block-level U.S. Census data.
A housing unit is defined as a house, apartment, mobile
home, or a room or group of rooms (U.S. Census Bureau
2002a, 2002b). Together the measures anthropogenic
land cover and housing units offered a more complete
picture of human influence because anthropogenic land
cover included all land covers that had a regular human
use associated with them and houses provided additional
detail that is often obscured in satellite images of forested
ecosystems (Radeloff et al. 2001) and hence is missed in
land-cover and -use maps. We did not analyze human population data because housing units and human population
were highly correlated (r > 0.97 in the study area). Both
databases were mapped in Albers conic equal area projection (NAD 83) and analyzed with ArcInfo geographic
information systems (GIS) (ESRI, Redlands, California)
and Erdas Imagine (Leica Geosystems LLC, Norcross,
Georgia).
Circular landscapes of approximately 1200 km2 centered on each BBS route were clipped from the humaninfluence databases following protocols in Flather and
Sauer (1996) and Donovan and Flather (2002). Under
Lepczyk et al.
409
these protocols, some landscapes may overlap if the BBS
routes are fairly close to one another. In each landscape
we calculated the total amount of anthropogenic land
cover and total number of housing units as measures of
human influence. Total anthropogenic land cover was
described as a proportion of the total amount of terrestrial surface area within each landscape scene, and housing units were log 10 -transformed to meet assumptions of
normality.
Modeling Biodiversity and Human Influence
We established a framework of 12 candidate models that
were suggested by the 3 hypotheses (Table 1). These
models considered the effects of anthropogenic land
cover and housing units separately and together on native species, from both linear and quadratic perspectives.
The models applied a correlative approach similar to Currie’s (1991:28) who said, “Even though correlations do
not demonstrate causality, they do serve 2 useful functions. When a correlation predicted by a hypothesis is
not observed, the hypothesis may be considered false.
Furthermore, when a correlation predicted by a hypothesis is weaker than other observed correlations, one may
conclude a better hypothesis exits.”
We used regression analysis to test the 12 candidate
models (Table 1) by modeling 3-year mean species richness and abundances against the proportion of anthropogenic land cover, the log 10 of total number of housing units, and their interaction. Abundance models were
deemed significant if the adjusted R2 ≥ 0.1 and the model
F statistic exceeded the p ≤ 0.05 threshold. If significant,
Table 1. Models used to test the 3 hypothesesa explaining human
influence on biodiversity.b
Model
1
2
3
4
5
6
7
8
9
10
11
12
Model descriptionc
dependent = land
dependent = houses
dependent = land + houses
dependent = land + houses + (land × houses)
dependent = land + land2
dependent = houses + houses2
dependent = land + land2 + houses
dependent = land + land2 + houses + (land
× houses)
dependent = land + houses + houses2
dependent = land + houses + houses2 + (land
× houses)
dependent = land + land2 + houses + houses2
dependent = land + land2 + houses + houses2
+ (land × houses)
a Productivity, ecosystem-stress, and intermediate-disturbance hypotheses.
b The 12 models were run separately for the dependent variables of
native species richness and abundance of each species.
c Key: land, the proportion of anthropogenic land cover; houses, the
log 10 of total number of housing units.
determination of the “best” relationship from among the
competing models of the same dependent variable was
based on Akaike’s information criterion (AIC; Burnham
& Anderson 2002). Nevertheless, because AIC can overparameterize models by adding variables that are not significant and do not improve fit (Guthery et al. 2005), we
selected the simplest model for each species abundance
that was within AIC of 2 of the minimum AIC model and
had significant model parameters. For species richness
we fit all 12 candidate models and used the AIC model
weights to define a confidence set of candidate models
(i.e., those models with AIC weights within 10% of the
highest weight [Royall 1997]). This confidence model
set and their associated weights were used to estimate a
model-averaged prediction of the richness of native bird
species for each landscape.
Because of the potential for spatial autocorrelation,
we conducted a subsequent analysis of the best models.
Specifically, we examined our data for spatial autocorrelation by first modeling the dependent variable (i.e.,
richness or abundance) against latitude and longitude
and saving the residuals. We then modeled the residuals
against the human-influence variables that were considered significant in the original model to see whether they
remained significant (p ≤ 0.05) and whether the model’s
R2 was still >0.1. For instance, White-eyed Vireo (Vireo
griseus) abundance was best described initially by anthropogenic land cover alone (model 1; see Results). We
then ran a second model of White-eyed Vireo abundance
against latitude and longitude and saved the residuals. In
turn, we modeled these residuals against anthropogenic
land cover, which still yielded a significant fit (adjusted
R2 = 0.23, p = 0.00008) after the influences of latitude
and longitude were removed. Because the richness model
and abundance models of a random sample of 20 species
all yielded significant fits following this process of testing for spatial autocorrelation, we report only the original model fits. We report both the adjusted R2 (presented hereafter as adj. R2 ) and AIC for all 12 models of
species richness, but only the top models and AIC of
the second-best model for species abundances.
All best models meeting our selection criteria were visually inspected on the basis of their scatter plots and
regression equations to assign which hypothesis (Fig.
1) they supported. Simple linear models were easily assigned to either productivity (+) or ecosystem-stress (−)
hypotheses, whereas quadratic models were assigned as
either positive (+), negative (−), or intermediate (I), depending on where the apex of the parabola occurred
(Fig. 3). Thus, a quadratic model could be essentially
positive (supporting the productivity hypothesis) or negative (supporting the ecosystem-stress hypothesis) and
not only intermediate. For models in which both anthropogenic land cover and housing units were significant but
had opposite or differing influences (e.g., positive for anthropogenic land cover and negative for housing units),
Conservation Biology
Volume 22, No. 2, 2008
Human Impacts on Diversity
410
III
II
Species diversity
I
Human influence
Figure 3. Illustration of how quadratic models relate
to the productivity, intermediate-disturbance, and
ecosystem-stress hypotheses. If the parabola’s apex
occurs at the x-axis (I), where human influence is the
lowest, the relationship is an ecosystem-stress
relationship. If the parabola’s apex occurs at or near
the greatest level of human influence (II), the
relationship is a productivity relationship. If the
parabola’s apex occurs in the mid ranges of human
influence (III), the relationship is an
intermediate-disturbance relationship.
conclusive assignment to one of any of the 3 hypotheses
could not be made.
Relationship of Natural-History Factors
with Human Influence
We also conducted a separate natural-history analysis
to discern the factors that may portend the nature of
species’ responses to human influence and to not limit
ourselves to testing an a priori set of expectations. In
other words, we grouped species by common response
and then looked for patterns in shared natural-history attributes, instead of using a guild approach that assumes
members of the guild respond similarly to a stressor
(sensu Severinghaus 1981), thereby providing information on what species were responding.
We initially tested for relationships between naturalhistory classes (body mass, fledging, incubation, nest
height, nest type, clutch size, mode of foraging, and mode
of development) and human influence with a multivariate approach, whereby we grouped all species that exhibited the same type of relationship (i.e., positive, negative,
mixed [+/− or −/+ with the 2 human-influence factors]
or negative quadratic). Nevertheless, with the exception
Conservation Biology
Volume 22, No. 2, 2008
of foraging, there were no significant relationships between natural-history classification and direction of the
relationship. Because of this lack of difference, we conducted a second analysis, whereby we pooled all species
exhibiting a relationship with human influence and compared them with species exhibiting no relationship for
each natural-history classification.
To test for the possibility of phylogenetic dependence
in the natural-history data of individual species (Harvey
& Pagel 1991), we conducted within-family contrasts for
each natural-history attribute following the approach of
Norris and Harper (2004). For each family we calculated
the difference in the mean natural-history attribute between the uncorrelated and correlated species. Hence,
a positive value would indicate that the uncorrelated
species had a greater value for that natural-history attribute (e.g., greater mass or nest height from ground).
We then tested whether the difference score was consistently positive or negative across families with a 1-tailed
t test and a null hypothesis of zero (Norris & Harper
2004). Notably, the t test could only be run across 14 of
the 38 total families that contained at least one species in
both uncorrelated and correlated classes. Because none
of the within-family contrasts were significant (i.e., p >
0.05), thus indicating no phylogenetic dependence in the
analyses, we present only the results based on the initial
comparison. Natural-history determinations were from
Dunning (1992), Ehrlich et al. (1988), and Poole (2005)
(see Supplementary Material). All statistical analyses were
performed with Systat 10 (SPSS 2000).
Results
The model with the highest weight of evidence indicated
that richness of native birds exhibited a negative relationship with anthropogenic land cover, an intermediate
relationship with the number of housing units, and an
interaction between anthropogenic land cover and housing (model 12, Table 2). The same model, but without the
interaction term (model 11, Table 2), and a land-coveronly model (model 5, Table 2) were close competitors
of the best model. These 3 models comprised the confidence set based on Royall’s (1997) suggested criterion.
Under this criterion, model 7, and all remaining candidate models, were excluded from the confidence set because they were >13 times (based on the ratio of model
weights [0.66/0.05]) less likely to be the best explanation for avian diversity when compared with model 12
(Table 2).
The model-averaged prediction among the confidence
set (Fig. 4a) indicated that species richness decreased
with increasing levels of anthropogenic land cover,
but exhibited an intermediate relationship with housing units. At very low levels of housing (<1000 units),
Lepczyk et al.
∗ Model number and coefficients are described in Table 1. Key: -2LL, -2log likelihood; w , AIC weights. The w was recalculated for only the 3 models in the confidence set, yielding weights of
i
i
0.125 (model 5), 0.153 (model 11), and 0.722 (model 12). These values were used to determine the predicted richness in Fig. 4.
0.025
0.004
0.0005
0.01
0.008
0.045
0.003
0.31
0.001
0.001
0.55
0.45
3254.7
0.001
<.0001
<.0001
3422.3 5 × 10−40
3256.5 0.0006
<.0001 0.679
3258
0.0003
0.27
0.386
3245.9
0.11
0.70
3424.4 2 × 10−40
0.578
3247.6
0.05
0.879
0.576
3249.3
0.02
0.558
0.437
3251.9
0.006 <.0001 0.010
3252.9
0.004
0.012
0.006
3245.6
0.14
0.546
0.054
3242.6
0.66
0.023
0.008
3254.6 12.3
3422.3 180
3256.4 14.1
3257.8 15.5
3245.8
3.5
3424.3 182
3247.5
5.2
3249.1
6.8
3251.8
9.5
3252.7 10.4
3245.4
3.1
3242.3
0
3
3
4
5
4
4
5
6
5
6
6
7
3248.6
3416.3
3248.4
3247.8
3237.8
3416.3
3237.5
3237.1
3241.8
3240.7
3233.4
3228.3
0.366
0.044
0.365
0.364
0.381
0.042
0.380
0.379
0.374
0.374
0.385
0.391
.
.
.
.
.
0.22
.
.
−4.19
−5.47
−3.28
−6.00
.
.
.
.
−28.62
.
−30.09
−32.47
.
.
−26.59
−34.63
.
.
.
−3.54
.
.
.
3.01
.
5.96
.
13.88
.
−7.24
0.60
2.81
.
−9.05
−0.83
−2.82
35.20
42.10
26.49
39.91
−32.51
.
−32.87
−19.47
−3.47
.
−1.48
−10.40
−33.97
−56.88
−6.00
−50.84
95.01
104.03
92.83
84.66
90.65
107.69
93.46
100.46
23.04
15.37
38.62
25.50
1
2
3
4
5
6
7
8
9
10
11
12
houses interaction
houses
Model intercept
land
houses interaction
land
2
Coefficients
Table 2. Native species richness results for each possible model in Table 1.∗
2
adj. R
2
-2LL
K
AIC
Δ AIC
AICC
wi
land
p
land 2
0.001
0.91
houses2
411
species richness quickly declined with increasing levels of anthropogenic land cover (Fig. 4a). On the other
hand, as the number of housing units on the landscape
increased, the slope of the relationship between richness
and anthropogenic land cover decreased and ultimately
began to show a slight quadratic relationship when housing numbers reached 1,000,000 units. This quadratic relationship at high housing density illustrates the interaction
between housing units and anthropogenic land cover,
but the relationship was tenuous as evidenced by only
a slight increase in species richness (<4 species) when
<50% of the landscape was anthropogenic (Fig. 4b). Nevertheless, there were few landscapes where either very
low or very high numbers of housing units (Fig. 4c) existed. Regardless of how many housing units were present
on the landscape, the relationship between species richness and the number of housing units always exhibited
a negative quadratic relationship (i.e., intermediate; Fig.
4d) that became more pronounced as anthropogenic land
cover increased.
Native species abundances were also strongly correlated with the amount of human influence present in the
landscape. Overall, 62.9% of all species (83 of 132; see
Supplementary Material) were related to either one or
both measures of human influence (maximum adj. R2 =
0.55; mean of adj. R2 = 0.23), well above the level expected due to random chance (7 significant correlations
would be expected by chance at p = 0.05). In 33 of
these relationships, abundance decreased with increasing levels of anthropogenic land use and housing density measured either alone or together (Table 3). Only
5 of the species exhibited a positive correlation with a
model containing a solitary human-influence measure or
both together. Similarly, only 6 species exhibited negative quadratic relationships that fit the characteristics
of intermediate disturbance (curve III, Fig. 3) with solitary or combined human-influence models. The remaining 39 species fell into 1 of 6 possible relationships (Table
3), whereby species exhibited either mixed relationships
with the 2 human-influence measures (i.e., positive relationship with one measure and negative relationship
with the second measure) or an intermediate relationship with 1 human-influence measure and a positive or
negative relationship with the second measure.
Among the 83 species that had significant relationships
with human influence, there were large differences in
terms of which models were selected (Table 3). For instance, model 6 (houses as a quadratic relationship) was
never selected, whereas model 7 (quadratic relationship
of land cover + houses) was selected 19 times. Among the
models that contained only one human-influence measure, land cover (models 1 and 5) was selected 19 times
compared with 4 times for houses (models 2 and 6).
Model 12 (best model for predicting native bird richness)
was selected as the best model predicting bird abundance
for only 5 species.
Conservation Biology
Volume 22, No. 2, 2008
Human Impacts on Diversity
412
Figure 4. (a) Predicted
native species richness in
relation to the proportion of
anthropogenic land cover
and the log 10 total number
of housing units on the
landscape. (b) Predicted
(surface) and observed (•;
i.e., Comdyn estimates)
species richness in relation
to anthropogenic land
cover. (c) Location of
landscapes on the
anthropogenic land-cover
and housing-unit gradient
(log scale). (d) Predicted
(surface) and observed (•;
i.e., Comdyn estimates)
species richness in relation
to log 10 total housing units.
Natural-history traits poorly predicted whether a bird
species was positively or negatively correlated with human influence. The only exception was foraging, in
which a greater proportion of terrestrial foragers (i.e.,
birds feeding on land for one or several food types, such
as granivores and frugivores) had a negative response
to human influence (χ2 = 17.4, df = 9, p = 0.048).
Nevertheless, natural-history traits predicted well species
that exhibited a relationship with human influence (i.e.,
species represented in Table 3) compared with those that
exhibited no relationship. Birds correlated with human
influence were significantly smaller in body mass, weighing only 7.2% as much as species not related to human
influence (mean [SE] 45.6 g [7.7] vs. 636.4 g [173.1];
t 130 = 4.44, p < 0.0001), and they fledged in about half
the time required by species that were unaffected by human influence (13.8 days [0.7] vs. 27.3 days [2.7]; t 128 =
6.03, p < 0.0001). Incubation times followed the pattern
of fledging times in that species correlated with human
influences hatched nearly 33% sooner than unaffected
species (13.4 days [0.4] vs. 19.6 days [1.1]; t 130 = 6.26,
p < 0.0001). Nevertheless, clutch sizes were similar for
correlated species versus uncorrelated species (4.4 eggs
[0.2] vs. 4.6 eggs [0.3]; t 130 = 0.58, p = 0.82). Corre-
Conservation Biology
Volume 22, No. 2, 2008
lated species more often exhibited altricial development
(67.5%), whereas uncorrelated species were more likely
(73.3%) to exhibit precocial development (χ2 = 9.51,
df = 1, p = 0.002). Species responding to human influences had minimum nest heights 42% closer to the
ground than species exhibiting no relationship (3.3 m
[0.6] vs. 7.8 m [1.4]; t 129 = 3.52, p = 0.0006) and were
predominately cavity- or cup-nesting species compared
with species unrelated to human influence (χ2 = 9.8,
df = 2, p < 0.007). Only a single species responding to
human influence was an aquatic forager and one other a
predator; all other species were generalists or terrestrial
foragers (χ2 = 25.4, df = 3, p < 0.0001).
Discussion
Overall, the pattern of declining native species richness with increasing anthropogenic land cover and a
preponderance of declining species abundances with
both human-influence measures are consistent with the
ecosystem-stress hypothesis (Rapport et al. 1985). Although we found no evidence that supported the productivity hypothesis for species richness (Table 2), we
Lepczyk et al.
413
Table 3. Relationships between human influences and bird abundances for native bird species with significant relationships.
Direction of relationship (land, houses)a
Model
Times model selected
1
2
3
4
5
6
7
8
9
10
11
12
Total
5
4
3
9
14
0
19
8
7
3
6
5
83
+,+
−,−
2
2
5b
4b
3
4
8
1
2
5
1
1
+,−
−,+
1
2
I,+
33
+,I
−,I
I,I
4b
6
1
1
2
2
2
3
1
10
5
1
3
1
1
5
I,−
9
1
4
1
2
9
3
1
6
1
1
6
a Direction
of relationship indicates whether land cover (land) and housing units (houses) were positively (+), negatively (−), or
intermediately (I) related to bird abundance.
b Only one independent variable included in model results.
did find support for the intermediate-disturbance relationship. Specifically, richness of native species exhibited a negative quadratic relationship with housing units
and the interaction of housing units and anthropogenic
land cover (Fig. 4a).
Of the 83 species abundances that exhibited a relationship, the plurality (33) were negative (Table 3, Supplementary Information). Nearly all 33 species that exhibited negative correlations require specific habitat types
that are being reduced through human modification (e.g.,
wetland drainage, simplification of forest structure, loss
of shrub, and ground cover), which can also increase
nest predation. For instance, 12 of the negatively affected species were warblers, a group of species often restricted to large intact forests. On the other hand, the 5
species that were positively correlated with human influence were those that have cosmopolitan distributions
(e.g., American Robin [Turdus migratorius]) and tend
to be frequent users of human subsidies (e.g., House
Wren [Troglodytes aedon], Mourning Dove [Zenaida
macroura]) such as nest boxes and supplemental food
(Lepczyk et al. 2004). The 6 species that exhibited purely
intermediate-disturbance relationships tended to be either edge or shrub species (e.g., Eastern Towhee [Pipilo
erythrophthalmus]) or human commensals (e.g., American Crow [Corvus brachyrhynchos]; Marzluff 2001). Furthermore, 15 species had a negative relationship with
one human-influence measure and an intermediate relationship with the other. Regardless of which factor was
negative and which was intermediate, the species exhibiting this combination of relationships were dominated by
woodpeckers (e.g., Hairy Woodpecker [Picoides villosus]) and grassland/wet-meadow species (e.g., Savannah
Sparrow [Passerculus sandwichensis], Common Yellowthroat [Geothlypis trichas]). Overall, though, the de-
crease of most native species abundances coupled with
very few exclusively positive or negative quadratic relationships provided further support for the ecosystemstress hypothesis.
Approximately 23% (19/83) of the species exhibited
a relationship with human influence that was best explained by anthropogenic land cover alone (i.e., models 1
and 5) rather than by housing units alone (approximately
5% [4/83]) (i.e., models 2 and 6). The second-best models
added only one additional species that was best explained
by model 1. Interestingly, however, the majority (72%)
of species abundances were best explained by a combination of both measures of human influence and their
interaction (Table 3; Supplementary Material). Thus, anthropogenic land cover was an important predictor by
itself in a large proportion of species, but model performance was enhanced in most instances by including
housing unit information. Aside from enhancing model
performance, housing units were highly correlated with
human population (r > 0.97) and thus may serve as a
useful surrogate for how human population relates to
biodiversity.
Although natural-history traits did not predict whether
or not a species was positively or negatively correlated
with human influence, they did predict well species that
had a relationship with human influence. That is, naturalhistory traits predicted the species that covaried with
human influence, but not how they covaried. Considering the natural-history factors together, the species that
correlated either positively or negatively with humaninfluence measures tended to be similar to altricial passerine birds.
Bird communities are frequently used as indicators of
environmental quality (Mayer & Cameron 2003) and are
thought to be a useful proxy for assessing the impact of
Conservation Biology
Volume 22, No. 2, 2008
414
human influence on biodiversity (Balmford & Long 1995;
Garson et al. 2002). We found that species richness and
abundances were indeed closely related to the degree
of human influence and that for the majority of species
neither human land use nor housing development was
beneficial. In terms of diversity the overall relationship
indicated that species richness was greatest where anthropogenic land cover was lowest and the number of
housing units was moderate (Fig. 4a). Similarly, for the
majority of species abundances, their relationships with
human influence were either negative or negative and
intermediate. Thus, the predominant patterns displayed
by the richness and abundance relationships were consistent with the predicted effects of stressed ecosystems
(Rapport et al. 1985) and to a lesser degree with the
effects of disturbance regimes on species richness.
Although most of the relationships supported the
ecosystem-stress hypothesis, land cover and housing interacted (Fig. 4), resulting in an intermediate disturbance relationship for species richness. Specifically, this
dual support for intermediate-disturbance and ecosystemstress hypotheses varied depending where on the model
predicted surface of houses and land cover the landscape
was located (Fig. 4a) and may be due to several factors.
For instance, because we classified agricultural lands as
anthropogenic, our results may be an artifact of the low
housing density on predominately anthropogenic landscapes. Although possible, this explanation is unlikely
because there were no landscapes that had <1000 housing units and were more than 50% anthropogenic (Fig.
4c). On the other hand, because landowners often manage and subsidize their property for wildlife (Lepczyk et
al. 2004), houses may be serving as a proxy for landowner
activities that cause elevated species richness in low to
moderately populated landscapes. In other words, as the
human population increases, there is a concordant increase in the number of houses, which in turn leads to increased habitat manipulation and resource diversity. This
increased resource diversity results in increased richness
up to a moderate level (1,000–10,000 units) of housing
(Fig. 4d). After moderate levels are exceeded, the landscape becomes too disturbed (Fig. 4c), even if manipulation continues, which leads to lower species richness.
A case in point is that suburban landowners (i.e., low to
intermediate housing density) plant vegetation on their
property intentionally for birds in greater proportion than
landowners in rural or urban locations (Lepczyk et al.
2004), thereby providing increased resource diversity at
intermediate housing levels. Whatever the case may be,
the relationships between housing units and land cover
suggest that there is an interesting and complex relationship between them.
Our findings suggest that the positive correlation between human population density and biodiversity observed in other studies (Balmford et al. 2001; Gaston &
Evans 2004; Luck et al. 2004) is likely related to a pattern
Conservation Biology
Volume 22, No. 2, 2008
Human Impacts on Diversity
of spatial congruence that appears at broad spatial scales.
Continental-scale studies with geographically large observation units capture extremes in ecosystem productivity to which biodiversity and human-population density
respond similarly. Furthermore, researchers using range
maps or data that are not spatiotemporally matched to
human population data may find differing relationships
compared with spatiotemporally matched data. In contrast, although our study was conducted at a regional
scale, it did include a productivity gradient (Hurlbert &
Haskell 2003), and we used a finer resolution of analysis that was spatiotemporally matched. Moreover, the
direction of relationship between richness and human influence appears to be strongly correlated with the scale
of analysis (Pautasso 2007). Thus, our failure to find evidence that supports the productivity hypothesis does not
mean birds (richness or abundance) do not positively covary with productivity. It could be that human uses of
the land do not covary as predicted by the productivity hypothesis. Because we did not measure productivity
directly (e.g., NDVI), but chose to test the productivity hypothesis implicitly using 2 measures of human influence, the direct relationship between productivity and
bird communities in this region of the country awaits future research.
Several caveats to our findings need to be considered.
First, it is possible that a positive relationship between
species richness and human influence occurs when humans first settle an area but that this relationship disappears over time as species become locally extirpated,
which results in a negative relationship. Unfortunately,
existing data do not allow for examination of changes
in the relationship of species richness and human influence over time. Second, some species (e.g., waterfowl,
shorebirds) may not have exhibited a relationship to human influence because they were poorly censused by
the BBS methodology. Third, even when species can be
easily censused, they may be rare, making strong relationships difficult to detect. Fourth, some anthropogenic
land cover classes (e.g., pasture/hay) provides suitable
breeding habitat and are managed in a way that permits successful reproduction. As a result, some anthropogenic land cover classes would perhaps be more accurately described as seminatural and may actually be
more similar to what we had classified as natural land
cover.
Our results support a 2-pronged approach to conservation. In areas with little or no human influence, ecosystems should be preserved or managed in ways that discourage human development (e.g., acquisition, conservation easements, economic incentives). Such preemptive
conservation is pragmatic and cost-effective in comparison with efforts to save or restore degraded or endangered species and landscapes (Norris & Harper 2004).
On the other hand, conservation priorities should not
be limited to pristine landscapes because our results
Lepczyk et al.
suggest that avian richness and the relative abundance
were estimated to peak at intermediate housing densities
or anthropogenic land cover, depending on overall landscape composition. In human-dominated areas, land managers should consider reestablishment of natural and seminatural habitats (McKinney 2002) when the expected
biodiversity gain justifies the greater cost associated with
restoration. Such an approach is especially important in
relatively pristine locations given that the U.S. population
is expected to increase by 65 million people between
1995 and 2025 (Fischer & Heilig 1997). Given the historical trend toward decentralization of human populations,
resulting in greater suburban and rural sprawl (Hammer
et al. 2004; Radeloff et al. 2005; Lepczyk et al. 2007),
our results presage an expanding human influence on
avian diversity. As they apply in human-dominated areas,
our results highlight the importance of extending conservation efforts to private lands and integrating ecological
principles in land-use planning at broad spatial scales.
Supplementary Material
A list of species investigated, a description of the bestfit models, and the natural-history traits of species are
available as part of the on-line article from http://www.
blackwell-synergy.com/ (Appendix S1). The author is responsible for the content and functionality of these materials. Queries (other than absence of the material) should
be directed to the corresponding author.
Acknowledgments
We thank J. Fallon at the U.S. Geological Survey Patuxent
Wildlife Research Center for BBS data and maps and M.
Linderman, M. Knowles, and T. Hawbaker for assistance
with data processing. We are grateful to J. Brashares and
R. King for insightful comments and reviewing a draft of
the manuscript and to T. Donovan and 3 anonymous reviewers who helped improve the manuscript. Financial
support for this research was provided by the Michigan
Agricultural Experiment Station, Michigan State University College of Social Science, National Science Foundation (NSF) (216450 and 709717), NSF Interdisciplinary
Informatics Fellowship (DBI-0306078), U.S. Department
of Agriculture Forest Service, North Central Research Station and Rocky Mountain Research Station, and U.S. Environmental Protection Agency STAR Fellowship program
(U-91580101-0).
Literature Cited
Balmford, A., and A. Long. 1995. Across-country analyses of biodiversity
congruence and current conservation effort in the tropics. Conservation Biology 9:1539–1547.
415
Balmford, A., J. L. Moore, T. Brooks, N. Burgess, L. A. Hansen, P.
Williams, and C. Rahbek. 2001. Conservation conflicts across Africa.
Science 291:2616–2619.
Beissinger, S. R., and D. R. Osborne. 1982. Effects of urbanization on
avian community organization. Condor 84:75–83.
Blair, R. B. 1996. Land use and avian species diversity along an urban
gradient. Ecological Applications 6:506–519.
Boulinier, T., J. D. Nichols, J. R. Sauer, J. E. Hines, and K. H. Pollock.
1998. Estimating species richness: the importance of heterogeneity
in species detectability. Ecology 79:1018–1028.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd
edition. Springer-Verlag, New York.
Burnham, K. P., and W. S. Overton. 1979. Robust estimation of population size when capture probabilities vary among animals. Ecology
60:927–936.
Bystrak, D. 1981. The North American breeding bird survey. Studies in
Avian Biology 6:34–41.
Cam, E., J. D. Nichols, J. R. Sauer, J. E. Hines, and C. H. Flather.
2000. Relative species richness and community completeness: birds
and urbanization in the Mid-Atlantic states. Ecological Applications
10:1196–1210.
Carroll, C., R. F. Noss, P. C. Paquet, and N. H. Schumaker. 2004. Extinction debt of protected areas in developing landscapes. Conservation
Biology 18:1110–1120.
Chown, S. L., B. J. van Rensburg, K. J. Gaston, A. S. L. Rodrigues, and A.
S. van Jaarsveld. 2003. Energy, species richness, and human population size: conservation implications at a national scale. Ecological
Applications 13:1233–1241.
Clergeau, P., J. L. Savard, G. Mennechez, and G. Falardeau. 1998. Bird
abundance and diversity along an urban-rural gradient: a comparative study between two cities on different continents. Condor
100:413–425.
Connell, J.H. 1978. Diversity in tropical rain forests and coral reefs.
Science 199:1302–1310.
Donovan, T. M., and C. H. Flather. 2002. Relationships among North
American songbird trends, habitat fragmentation, and landscape occupancy. Ecological Applications 12:364–374.
Dunning, J. B. 1992. CRC handbook of avian body masses. CRC Press,
Boca Raton, Florida.
Ehrlich, P. R., D. S. Dobkin, and D. Wheye. 1988. The birder’s handbook:
a field guide to the natural history of North American Birds. Simon
& Schuster, New York.
Fischer, G., and G. K. Heilig. 1997. Population momentum and the
demand on land and water resources. Philosophical Transaction of
the Royal Society of London, B–Biological Science 352:869–888.
Flather, C. H., and J. R. Sauer. 1996. Using landscape ecology to test
hypotheses about large scale abundance patterns in migratory birds.
Ecology 77:28–35.
Foley, J. A., et al. 2005. Global consequences of land use. Science
309:570–574.
Garson, J., A. Aggarwal, and S. Sarkar. 2002. Birds as surrogates for
biodiversity: an analysis of a data set from Southern Quebec. Journal
of Bioscience 27:347–360.
Gaston, K. J. 2005. Biodiversity and extinction: species and people.
Progress in Physical Geography 29:239–247.
Gaston, K. J., and K. L. Evans. 2004. Birds and people in Europe. Proceedings of the Royal Society London B 271:1649–1655.
Genet, K. S., C. A. Lepczyk, R. Christoffel, L. G. Sargent, and T. R. Burton.
2008. Using volunteer monitoring programs for anuran conservation
along a rural-urban gradient in Southern Michigan USA. (In press)
in J. C. Mitchell and R. E. J. Brown, editors. Urban Herpetology.
Herpetological Conservation. Volume 3. Society for the Study of
Amphibians and Reptiles, Salt Lake City, Utah.
Guthery, F. S., L. A. Brennan, M. J. Peterson, and J. J. Lusk. 2005. Information theory in wildlife science: critique and viewpoint. Journal
of Wildlife Management 69:457–465.
Conservation Biology
Volume 22, No. 2, 2008
416
Hammer, R. B., S. I. Stewart, R. L. Winkler, V. C. Radeloff, and P. R.
Voss. 2004. Characterizing dynamic spatial and temporal residential
density patterns from 1940–1990 across the North Central United
States. Landscape and Urban Planning 69:183–199.
Harvey, P. H., and M. D. Pagel. 1991. The comparative method in evolutionary biology. Oxford University Press, Oxford, United Kingdom.
Hawkins, B. A., et al. 2003. Energy, water, and broad-scale geographic
patterns of species richness. Ecology 84:3105–3117.
Hines, J. E., T. Boulinier, J. D. Nichols, J. S. Sauer, and K. H. Pollock.
1999. COMDYN: software to study the dynamics of animal communities using a capture-recapture approach. Bird Study 46:S209–S217.
Hope, D., C. Gries, W. Zhu, W. F. Fagan, C. L. Redman, N. B. Grimm, A. L.
Nelson, C. Martin, and A. Kinzig. 2003. Socioeconomics drive urban
plant diversity. Proceedings of the National Academy of Sciences of
the United States of America 100:8788–8792.
Hurlbert, A. H., and J. P. Haskell. 2003. The effect of energy and seasonality on avian species richnes and community composition. The
American Naturalist 161:83–97.
Lepczyk, C. A., A. G. Mertig, and J. Liu. 2004. Assessing landowner activities related to birds across rural-to-urban landscapes. Environmental
Management 33:110–125.
Lepczyk, C. A., R. B. Hammer, S. I. Stewart, and V. C. Radeloff. 2007.
Spatiotemporal dynamics of housing growth hotspots in the North
Central U.S. from 1940 to 2000. Landscape Ecology 22:939–952.
Luck, G. W., T. H. Ricketts, G. C. Daily, and M. Imhoff. 2004. Alleviating
spatial conflict between people and biodiversity. Proceedings of
the National Academy of Sciences of the United States of America
101:182–186.
McDonnell, M. J., and S. T. A. Pickett. 1990. Ecosystem structure and
function along urban rural gradients: an unexplained opportunity
for ecology. Ecological Applications 71:1232–1237.
McKinney, M. L. 2002. Urbanization, biodiversity, and conservation.
BioScience 52:883–890.
Marzluff, J. M. 2001. Worldwide urbanization and its effects on birds.
Pages 19–47 in J. M. Marzluff, R. B. Bowman, and R. Donnelly,
editors. Avian ecology and conservation in an urbanizing world.
Kluwer Academic Publishers, Boston.
Mayer, A. L., and G. N. Cameron. 2003. Landscape characteristics, spatial extent, and breeding bird diversity in Ohio, USA. Diversity and
Distributions 9:297–311.
Nichols, J. D., J. E. Hines, J. R. Sauer, F. W. Fallon, J. E. Fallon, and P.
J. Heglund. 2000. A double-observer approach for estimating detection probability and abundance from point counts. Auk 117:393–
408.
Norris, K., and N. Harper. 2004. Extinction processes in hot spots of
avian biodiversity and the targeting of pre-emptive conservation
action. Proceedings of the Royal Society London B 271:123–130.
Conservation Biology
Volume 22, No. 2, 2008
Human Impacts on Diversity
Pautasso, M. 2007. Scale dependence of the correlation between human population presence and vertebrate and plant species richness.
Ecology Letters 10:16–24.
Poole, A., editor. 2005. The birds of North American Online. Cornell Laboratory of Ornithology, Ithaca, New York. Available from
http://bna.birds.cornell.edu/BNA/ (accessed December 2004).
Radeloff, V. C., R. G. Hammer, P. R. Voss, A. E. Hagen, D. R. Field, and
D. J. Mladenoff. 2001. Human demographic trends and landscape
level forest management in the northwest Wisconsin Pine Barrens.
Forest Science 47:229–241.
Radeloff, V. C., R. G. Hammer, and S. I. Stewart. 2005. Rural and suburban sprawl in the U.S. Midwest from 1940 to 2000 and its relation
to forest fragmentation. Conservation Biology 19:793–805.
Rapport, D. J., H. A. Regier, and T. C. Hutchinson. 1985. Ecosystem
behavior under stress. The American Naturalist 125:617–640.
Royall, R. M. 1997. Statistical evidence: a likelihood paradigm. Chapman
& Hall, London.
Ricketts, T., and M. Imhoff. 2003. Biodiversity, urban areas, and agriculture: locating priority ecoregions for conservation. Conservation
Ecology 8: http://www.consecol.org/vol8/iss2/art1.
Robbins, C. S., D. Bystrak, and P. H. Geissler. 1986. The breeding bird
survey: its first fifteen years, 1965–1979. Resource publication 157.
U.S. Fish and Wildlife Service, Washington, D.C.
Sauer, J. R., B. G. Peterjohn, and W. A. Link. 1994. Observer differences
in the North-American breeding bird survey. Auk 111:50–62.
Sauer, J. R., J. E. Fallon, and R. Johnson. 2003. Use of the North American
Breeding Bird Survey data to estimate population change for bird
conservation regions. Journal of Wildlife Management 67:317–326.
Severinghaus, W. D. 1981. Guild theory development as a mechanism
for assessing environmental impacts. Environmental Management
5:187–190.
Sinclair, A. R. E., S. A. R. Mduma, and P. Arcese. 2002. Protected areas biodiversity benchmarks for human impact: agriculture and
the Serengeti avifauna. Proceedings of the Royal Society London
B 269:2401–2405.
SPSS. 2000. Systat. Version 10. Systat Software, Inc., San Jose, California.
U.S. Census Bureau. 2002a. Census 2000 summary. File 3A technical
documentation. U.S. Census Bureau, Washington, D.C.
U.S Census Bureau. 2002b. Technical documentation. U.S. Census Bureau, Washington, D.C.
Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997.
Human domination of Earth’s Ecosystems. Science 277:494–499.
Vogelmann, J. E., S. M. Howard, L. Yang, C. R. Larson, B. K. Wylie, and
N. van Driel. 2001. Completion of the 1990s National Land Cover
Data Set for the conterminous United States from Landsat Thematic
Mapper data and ancillary data sources. Photographic Engineering
and Remote Sensing 67:650–662.